The present invention relates generally to the field of workload management, and more particularly to dynamic pool allocation in workload management.
Workload management is the process of controlling the access to system resources for the work executing on a set of computer resources (for example, a mainframe computer). On a typical set of computer resources, many different applications execute at the same time. Typically, work management attempts to accomplish objectives such as the following: (i) execute work with consistent execution times; and/or (ii) allow predictable access to databases. In some work management systems, a system administrator classifies work into service classes. The classification mechanism uses work attributes like transaction names, user identifications or program names which specific applications are known to use. In addition, the system administrator defines goals and importance levels for the service classes representing the application work. The goals define performance expectations for the work. Goals can be expressed as response times, a relative speed (termed velocity) or as discretionary if no specific requirement exists.
A typical problem when dealing with workload management is to keep the expected deadlines for jobs that are critical for the business. For instance, a bank must be sure to process payment orders by a given date. It happens very frequently that critical jobs share the same resources as non-critical ones, for instance, the payment order process can share the same resources of another process for printing statistical reports that are not subject to any particular constraint in time. If the workload scheduling infrastructure caused the job to start according to the user's defined schedule, it could happen that the non-critical reporting process is scheduled to start earlier than the critical payment orders process (for instance, reporting is scheduled to start at 10.00 AM and order processing at 10.30 AM. In this case, the reporting process “steals” resources that could be better used to complete the order processing on time.